Mixed Rating Systems

Jump to navigation Jump to search

Mixed Rating Systems

You know, since there's a little bit of disagreement about what kind of rating system best reflects the kind of improvements valued in robots, I wonder if it would make sense to have a mixed system for overall ranking? What about using a condorcet voting method to resolve the rankings, where each ranking system is a "ballot" that fills out the rank of all robots? One "ballot" could be APS, another "ballot" could be W%, and perhaps another "ballot" could be S%. The resulting rankings would essentially be based on the consensus of the three component ranking systems. I think it's natural for robots that excel in all three metrics to rank high, and is perhaps a more universally agreed and more general notion of strength than just one metric alone.

Thoughts?

Rednaxela15:31, 21 September 2011

I like that approach myself. Though I think I would limit it to APS and W%.

Skotty16:30, 21 September 2011
 

Nevermind, forget what I said. I didn't read it close enough. I was thinking that I would like something that averages APS and W%, with the weight of each not necessarily being equal.

Skotty16:54, 21 September 2011
 

Averaged would be nice in that it would allow easier differentiation between small changes, but on the other hand, condorcet ranking has the property of not allowing one metric to dominate. I think there are advantages to each way of mixing.

Rednaxela17:56, 21 September 2011
 

Both condorcet and averaging can allow one kind of metric to dominate through clones. Similar methods together will overweight other unique methods. Ballots need to be independent from the candidates (methods) to avoid the clone problem.

MN05:53, 7 October 2011